While there is no network-wide policy on AI-generated answers such as those produced by ChatGPT, judging from our moderating experience over the past months and the related meta discussion, we the moderators now have decided to adopt a policy of banning AI-generated answers.

There are several reasons

  • users posting AI-generated content often do not seem able to check whether the AI-generated answer is accurate — moreover, they often do not interact with the OP or other users to clarify or improve the post
  • there is a high risk of flooding the site with plausible-sounding but incorrect answers.

Because of the adverse effects on the main goal of supporting a community producing high quality answers about the Japanese language, we will follow the policy on StackOverflow:

  • The use of ChatGPT or similar large language model AI to create posts on https://japanese.stackexchange.com/ is not permitted.
  • Posts created with ChatGPT or AI are subject to deletion without notice.
  • If a user is believed to have used ChatGPT after the posting of this policy, sanctions will be imposed to prevent them from continuing to post such content, even if the posts would otherwise be acceptable.
  • What about cases where I have no evidence that the poster has used a large language model like GPT-3 but I have a strong suspicion? Consider this new user’s posts: japanese.stackexchange.com/a/99124/816 The transcription error yaowarai is extremely unlikely to be an honest mistake because it’s not plausibly close to anything it could be confused for. However it is characteristic of the kind of hallucinations LLMs make. How should we proceed?
    – jogloran
    Mar 31 at 1:36
  • The poster in question then deleted their post after their odd mistake was called out, which really makes me suspicious.
    – jogloran
    Mar 31 at 3:04
  • 1
    @jogloran, I've got enough points to see deleted posts, and I see that a moderator deleted that, not the poster themselves -- likely due to the clear indications that the post was both AI-generated, and wrong. Mar 31 at 17:44
  • Got it. Hopefully this is enough negative reinforcement to prevent them from doing it again.
    – jogloran
    Mar 31 at 18:42


You must log in to answer this question.

Browse other questions tagged .